A typical audio teleconferencing system comprises a centralized facility or bridge and a plurality of participant terminals connected to the bridge. Audio message signals produced by the individual participant terminals are encrypted at the participant terminals and transmitted to the bridge in encrypted form. These message signals are received and decrypted by the bridge. The clear text message signals are then processed by the bridge, for example, the clear text message signals are summed. The resulting signal is then encrypted at the bridge and transmitted from the bridge to the participant terminals where decryption takes place. Illustratively, the bridge could just add the speech signals from all the participants in the teleconference and broadcast the sum in encrypted form. However, this is generally not done because this would also add the background noise from all the participants and would require unnecessarily large dynamic range. In general, the bridge adds the message signals from a subset of the participants in the teleconference.
One shortcoming of the type of teleconferencing system described above is that the message signals are present in the bridge in clear text decrypted form. Such bridges are therefore not suitable for secure teleconferencing. Accordingly, it is an object of the present invention to provide a secure audio teleconferencing system utilizing a bridge for bridging encrypted audio signals without first decrypting them so that clear text message signals are not present at the bridge.